Welcome to our community

Be a part of something great, join today!

  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

REDCODE wavelet codec advantages

Dennis S.

Well-known member
Joined
Aug 19, 2012
Messages
48
Reaction score
0
Points
0
Hi,

first i have to say, i am not a camera guy but i am in charge for our IT and as you all know, postproduction and IT, usually have a few things in common ;)
That being said, i just try to help our camera guys as good as i can with the little knowledge i have about the RED Epic, we purchased a while ago.

Our camera guys are currently filming abroad, mostly at 4K sometimes even 5K and are ordered to "trim" everything down as good as possible, so that we all make our life a little bit
easier when it comes to saving all the material on our main raid storage, at our production office.

Reviewing 4K/5K material on a macbook or a windows notebook isn`t a funny thing to do. We use a RED rocket for that purpose even on our main postproduction workstation but thats not an option for all of our notebooks/macbooks because we focus on nature documentary and its required that our camera people can review the files, while filming in the outbacks of this world ;)

As i was looking around the internet i found an interesting description of the REDCODE codec and the advantages wavelet based codecs can offer at wikipedia.
Here is the entire quote:

Wikipedia said:
However, since REDCODE is a wavelet codec, the files contain several lower resolution versions of the video; the codec uses each in sequence to build the next higher resolution version.
That means a 4K file can supply 2K, 1K, or even 0.5K footage directly without decoding the full 4K resolution data followed by scaling.

Thats sounds like the ideal solution for the problem i described above.
I am not sure if i understood it correctly but is it really possible to easily switch from a 4K recorded R3D file to lets say 2K so that proper playback on notebooks isn`t an issue, trim everything down to the scenes you really wanna have and translate the changes back to the 4K file immediately?
Seperate proxy files wouldn`t help that much i guess because you commit your changes to the proxy file and not to the 4K R3D you recorded which would make additional working steps necessary.
I am looking forward to your answers and would like to say thank you in advance for helping out an IT guy who is a little bit lost in the world of camera stuff. ^^

Dennis
 
Hi Dennis. The various resolution clips are generated at the moment only with the Red One. Not so with the Epic/scarlet. You can however create them in post through Redcine. But in all honesty, most laptops can playback the red footage at 1/4 or even half res pretty efficiently.
 
Hi Dennis. The various resolution clips are generated at the moment only with the Red One. Not so with the Epic/scarlet. You can however create them in post through Redcine. But in all honesty, most laptops can playback the red footage at 1/4 or even half res pretty efficiently.

I think you're referring to the QT pointer files of the RED One. And that is related to the wavelet and multiple resolutions, but is just a round-about way of interfacing them through QuickTime.

Wavelet compression of images, or video frames in this case, are a multi-tiered algorithm where you don't have to decompress them all the way to get at a useable image. In REDCINE-X, or other apps that properly support R3D, you pick your playback resolutions of full, 1/2, 1/4, 1/8... And those are based on how far you dig into the wavelet data. Choosing 1/8th resolution playback essentially only digs 1/8th of the way into the wavelet, only consumes 1/8th of the CPU horsepower needed to decode the full frame, delivers 1/8th of the resolution and runs 8 times faster. ...Or even perceptibly faster than that as 1/2 or full resolution on a given system may just be too much for that system to handle at all.

Current generation laptop computers using mobile i7 CPUs with adequate amounts of memory and respectable internal storage can play back 5K R3D files at 1/4 res in real-time without breaking a sweat. Which is all that's needed most of the time for any real-world playback and checking. Here and there, it can be stopped and increased to full resolution and viewed at 1:1 pixels if needed to verify a specific detail or critical focus, if necessary.

Proper use of wavelets essentially eliminates the need for offline workflows and rendered proxy formats. But it's still taking some time for most of the software out there to really capitalize on these advantages.
 
Thanks for all your fast replies, i appreciate it :)
So bascially it would be best to pick 1/2, 1/4, 1/8 or whatever the horsepower of the notebook is capable off, to
practically achieve the same, as you would when you generate QT wrappers from within REDCINE-X Pro or is there anything special,
less cpu dependant about the QT wrapped files?

In other words a 4K R3D file played back at 1/2 results in 2K with only half the CPU power necessary for the decoding and with twice the speed
so for example if i can`t play the 4K file in realtime (full) but maybe at 13fps, i can play it at 26fps with 1/2.
Is that correct?

Thank you :)
 
Thanks for all your fast replies, i appreciate it :)
So bascially it would be best to pick 1/2, 1/4, 1/8 or whatever the horsepower of the notebook is capable off, to
practically achieve the same, as you would when you generate QT wrappers from within REDCINE-X Pro or is there anything special,
less cpu dependant about the QT wrapped files?

The QT pointers, or QT wrapped files if you log them into FCP7, are nothing special. They're just flagged to run at a certain wavelet decode level. The wrappers and QT pointer files are just for the benefit of making R3D files play within QuickTime. You of course have to install the QuickTime R3D plugin, which is Mac only. It's also a dated and clunky way to work.

In other words a 4K R3D file played back at 1/2 results in 2K with only half the CPU power necessary for the decoding and with twice the speed
so for example if i can`t play the 4K file in realtime (full) but maybe at 13fps, i can play it at 26fps with 1/2.
Is that correct?

That's correct in a generalized, nut-shell way.

what aplication are you using for this??

Premiere does a fantabulous job. :) So does Sony Vegas Pro and now Edius 6.5. RedCine-X does alright, but it's optimized for use with the Rocket more so than being multi-threaded, so it's performance isn't all that great compared to other solutions out there.

I feel better in 4K wavelett than in 1080p AVCDHD. I LOVE Redcode.

Yep. Agreed. It's amazing how well the latest NLE software is handling R3D files. It's fun to see the looks on peoples' faces when I show them how well Premiere CS6 thrashes through 5K EPIC files and does it better, smoother and faster than FCP7 can deal with 1080p AVC from most consumer cameras. :)

Wish the epic still had QT proxies.

eh... why? The QT proxies always were, and still are, a very clunky and kludge way of working with R3D files. But if for some reason you still want them, you can generate them from within RC-X. Yes, even for EPIC 5K files. IMO, if you're still in a FCP7 editing environment, you're better off to generate QT LT proxies from within RC-X (or Premiere / AME, lol) and edit with those. The latest releases of the R3D plugin for Final Cut Studio has Rocket support if you do L&T ingest and transcode to ProRes. So that's one other way to ingest usable footage. FCP7 works best if you edit in one of the various flavors of ProRes, but Color has proper internal support for native QT-wrapped R3D, or you can re-link to the R3D's themselves on XML export for ingest into Resolve.

SpeedGrade also does a great job of handling R3D files natively and does many things snappier than Resolve does. Biggest complaint from most about SpeedGrade is monitoring as it requires an nVidia SDI card. That's changing soon though. ;) like IBC soon... ooops, and he walks away.. muttering something about cats and bags and whatnot.
 
Thank you guys.

@Jeff

You mentioned that Adobe Premiere 6 handles RED material really well.
Does Premiere CS6 really playback files that much faster compared to REDCINE-X (without a Rocket)?

Can i trim down the R3D files in Premiere and export them unchanged or do i need to export the files into another format
after i cut out some scenes for example.
 
If you want to trim R3D files and export the trimmed segments as R3D, you will have to use REDCINE-X for now. As for playback performance, Premiere is superior to REDCINE-X with and without a Rocket. However, we're comparing apples and oranges here when it comes to many tasks and there are many things REDCINE-X does that that Premiere does not and vice versa. You will often use them together.

You can always download Premiere and use it for the free trial period.
 
It would be nice to keep things simple for our camera guys because we all have a short time schedule for the whole project.
If i advice our camera guys to use Adobe Premiere for R3D playback and REDCINE-X for trimming the files they definitely will stare me down until i cry like a baby ^^

Honestly, what i would consider an ideal solution is a tool, which enables our guys to playback the 4K/5K Redfiles in realtime, (1/2 1/4 is acceptable of course) allow them to trim the files and export the trimmed sections afterwards
without loosing mich time for a transcoding process, so in other words export to R3D directly. The workflow to achieve that should be as simple as possible.
 
REDCINE-X is all you need for that playback / trim approach. Transcoding out of REDCINE-X runs great with a Rocket. If you don't have a Rocket, then you can transcode faster through Premiere / AME, if your system is up to competent specs.

Don't know what else to say besides that. But if you're trying to keep it as simple as possible and the least amount of tools and new software to cope with, then you'll want to go for REDCINE-X as the all around tool. REDCINE-X plays back just fine at 1/4 resolution on 2011 and later i7 Macbook Pro or equivalent spec'd PC laptops. 5KFF footage will often stutter in REDCINE-X on a system like that when at 1/4 speed and may need to be set to 1/8th. There are numerous factors in all of this too. The speed of the storage you read and write to for playback, offloading and transcodes is very important is the most common bottleneck I see in most every situation people come to me asking for advice.
 
eh... why? The QT proxies always were, and still are, a very clunky and kludge way of working with R3D files. But if for some reason you still want them, you can generate them from within RC-X. Yes, even for EPIC 5K files. IMO, if you're still in a FCP7 editing environment, you're better off to generate QT LT proxies from within RC-X (or Premiere / AME, lol) and edit with those. The latest releases of the R3D plugin for Final Cut Studio has Rocket support if you do L&T ingest and transcode to ProRes. So that's one other way to ingest usable footage. FCP7 works best if you edit in one of the various flavors of ProRes, but Color has proper internal support for native QT-wrapped R3D, or you can re-link to the R3D's themselves on XML export for ingest into Resolve.

SpeedGrade also does a great job of handling R3D files natively and does many things snappier than Resolve does. Biggest complaint from most about SpeedGrade is monitoring as it requires an nVidia SDI card. That's changing soon though. ;) like IBC soon... ooops, and he walks away.. muttering something about cats and bags and whatnot.

Not for me, For clients who want to view footage. Just to look and see what they got.
 
REDCINE-X is all you need for that playback / trim approach. Transcoding out of REDCINE-X runs great with a Rocket. If you don't have a Rocket, then you can transcode faster through Premiere / AME, if your system is up to competent specs.

I run a few tests with Premiere /AME to transcode trimmed files directly into DnxHD 36 our offline codec.

REDCINE-X plays back just fine at 1/4 resolution on 2011 and later i7 Macbook Pro or equivalent spec'd PC laptops.

Can`t confirm that. According to our camera guys it heavily depends on what kind of material is played back.
If for example the R3D contains a lot of movement then even 1/4 can`t be played back properly, at least thats what our camera people are reporting to me.
Of course the stuttering can be related to slow harddrive I/O. I might consider SSDs for our notebooks if i can confirm on that.

Nevvertheless, thank you very much for all your help
 
Can`t confirm that. According to our camera guys it heavily depends on what kind of material is played back.
If for example the R3D contains a lot of movement then even 1/4 can`t be played back properly, at least thats what our camera people are reporting to me.
AFAIK, amount of movement shouldn't matter with RED footage. Amount of detail in the image could possibly have a tiny impact on the actual bitrate, but I doubt if it would be enough to make a difference. Most likely none of the footage plays back properly, but it's just more apparent to the viewer when there's lots of movement.
 
Most likely none of the footage plays back properly, but it's just more apparent to the viewer when there's lots of movement.
+1

I would definitely take a look and see if there are other things that could be affecting performance of playback - like the footage being stored on a portable drive connected over USB2
 
Not for me, For clients who want to view footage. Just to look and see what they got.

As Jaakko already pointed out, use the RED player. It works a lot better and unlike the QT proxies, it's available on both Mac and Windows.

As for playback on the laptop at 1/4 resolution, it does vary with the format shot and the amount of detail in the shot. But mostly 1/4 is fine on any quad-core 2.3GHz or better system with enough RAM and fast enough storage. 5KFF and 5K 16:9 can still necessitate a drop to 1/8 on such a system at times. REDCODE is a variable-rate codec, so even when you choose 8:1 or 3:1, etc.. those are approximate ratios. More detail in a scene can mean that the compression isn't quite as effective. Higher compression ratios save disk space and impact quality. They also take more resources to decode. It's harder on a system to play back R3Ds recorded at 10:1 than that recorded at 7:1. Which puts a lot of guys in a bind without realizing it -- especially those with under-spec'd systems or tight budgets who are tempted to increase the compression ratio to save on storage.
 
Last edited:
Back
Top